Thoughts

Loose thoughts, jotted down as they come.

2026-05-05

Many software problems only appear through real use, and agent harnesses make that pattern obvious.

Software EngineeringAI AgentsHarnessEvaluationUX
2026-04-15

Embodied AI and CUA are the same thing in two different worlds — and the digital world model should be easier.

CUAEmbodied AIWorld Model
2026-04-14

CUA training infrastructure is three lines: Data, Eval, Train — connected by Gym.

CUATraining InfrastructureBenchmarks
2026-04-12

The missing product in the AI agent ecosystem isn't another framework — it's a client app designed for users.

AI AgentsProductInfrastructure
2026-04-10

In the age of AI-driven output explosion, promotion is no longer optional — it's a core skill.

Open SourcePromotionWorkflow
2026-04-04

Agent swarm communication is unsolved at the product level — markdown files and polling are still the default.

AI AgentsMulti-AgentSystem Design
2026-04-02

The bottleneck of long-horizon agentic tasks isn't prompting — it's that chat-trained models aren't built to act autonomously.

AI AgentsLLM TrainingAgentic AI
2026-03-25

CUA's capabilities are essentially three layers: thinking, discrete operations, and continuous operations.

CUAAI Agents
2026-03-19

Embodied AI and CUA share the same core challenges — world modeling, rule discovery, and experience memory.

AI AgentEmbodied AICUAWorld Model
2026-03-07

CLIzing everything — as agents grow stronger, more apps and websites should open interfaces for them.

AI Agents
2026-03-05

The fundamental difference between agentic and chat paradigms is the ability to explore through tool use.

AI Agents